Cooperative Spectrum Sensing based on Flatness in Spectral Domain under Noise Variance Uncertainty

We consider a cooperative spectrum sensing setup, based on the flatness measure in the spectrum of the received observations. This scheme extends the existing spectral flatness measure (SFM) detector in a multi-sensor scenario. Under Gaussian noise, the noise samples have a flattened spectrum in contrast to that of the primary signal samples and hence, the SFM under the null hypothesis is larger. We provide a detailed proof on deriving the asymptotic test statistic distribution and the asymptotically optimal detection threshold. This test is robust to the noise variance uncertainty. Through extensive Monte Carlo simulations and experimental results, we show that the proposed test performs better than the blind detector and the energy detector under practically relevant signal and fading models, all in the presence of noise variance uncertainty.